Search All of the Math Forum:
Views expressed in these public forums are not endorsed by
Drexel University or The Math Forum.



Re: Matlab  RMSvalue
Posted:
Oct 11, 2012 11:16 AM


"Tom Lane" <tlane@mathworks.com> wrote in message <k56js0$he1$1@newscl01ah.mathworks.com>... > > RMS1 = sqrt( mean(inVec)^2 + var(inVec) ); > > RMS2 = sqrt( 1/length(inVec) * sum(inVec.^2) ); > ... > > Why is there a difference ? Roundoff error or something else ? > > You're right, something else. Ordinarily var uses 1/(n1) rather than 1/n in > estimating the variance of a sample of size n. You can supply a flag to get > it to use 1/n, and then the results match: > > >> v = 1:10; > >> sqrt(mean(v)^2 + var(v)) > ans = > 6.2783 > >> sqrt(1/length(v) * sum(v.^2)) > ans = > 6.2048 > >> sqrt(mean(v)^2 + var(v,1)) > ans = > 6.2048 > >  Tom
There's also:
>> rmsval = norm(1:10) / sqrt(length(1:10))
rmsval =
6.2048e+000



